* src/search.c (Freplace_match): Don't test for change in search_regs
start and end, this is unreliable if change hooks modify text earlier
in the buffer (bug#35264).
* src/emacs.c (main): Call 'set_initial_minibuffer_mode' before
entering recursive-exit.
* src/minibuf.c (init_minibuf_once_for_pdumper): Don't call
'set_minibuffer_mode' here...
(set_initial_minibuffer_mode): ... set it in this new function.
(Bug#48446)
* src/lisp.h: Add prototype for 'set_initial_minibuffer_mode'.
Pgtk didn't update text and C-g didn't take effect while it is busy.
By timer interrupts, we can call pgtk_read_socket, and update texts
and handle C-g.
* src/pgtkterm.c (start_timer): New function to start timer.
(pgtk_show_hourglass): Remove code to start timer.
(pgtk_hide_hourglass): Remove code to stop timer.
(pgtk_term_init): Call start_timer().
* src/minibuf.c (read_minibuf): Call get_minibuffer before incrementing
minibuf_level, in case a hook function calls Factive_minibuffer_window.
(init_minibuf_once_for_pdumper): Create *Minibuf-0* here (moved from
init_minibuf_once), and set its mode, so that clicking in the mini-window
immediately after start up works (thanks, Eli Z.).
Also fix some unsafe coding.
* lisp/window.el (push-window-buffer-onto-prev): New function, extracted from
(record-window-buffer): Refactor by extracting the above, and removing the now
redundant parameter DO-MINIBUF.
* src/minibuf.c (zip_minibuffer_stacks, read_minibuf): Replace calls to
get_minibuffer (0) by nth_minibuffer (0). Replace calls to
Qrecord_window_buffer by calls to Qpush_window_buffer_onto_prev.
(Factive_minibuffer_window, read_minibuf_unwind): Call abort_emacs should an
"impossible" null value be returned by nth_minibuffer.
(read_minibuf): Move the get_minibuffer_call to just after the incrementation
of minibuf_level as a precaution against a missing buffer in
Vminibuffer_list.
(nth_minibuffer): Replace XCAR by Fcar, to allow (car nil) to work.
(init_minibuf_once): Create the inactive buffer *Minibuf-0*.
(syms_of_minibuf): New DEFSYM, Qpush_window_buffer_onto_prev.
* src/window.c (restore_window_configuration): Replace some XCARs and XCDRs
by Fcar_safe and Fcdr_safe.
On Wayland,
even if without WAYLAND_DISPLAY, --display DISPLAY works, but
gdk_display_get_name() always return "wayland-0", which may be
different from DISPLAY.
If with WAYLAND_DISPLAY, then it always returns WAYLAND_DISPLAY.
So pgtk emacs is confused and enter multi display environment.
To workaround this situation, treat all the wayland-* as the same
display.
* src/pgtkfns.c (is_wayland_display): New function to decide it is a
wayland connection.
(pgtk_display_info_for_name): If both are wayland connection, then
they are the same display.
* src/process.c (add_non_keyboard_read_fd): Make this a public function.
(add_process_read_fd): Fold old, static add_non_keyboard_read_fd guts
into here.
* src/xsmfns.c (ice_conn_watch_CB): Call add_non_keyboard_read_fd
(bug#43834).
'line-move-partial' should in general leave it to the display
engine to scroll or recenter the window due to vertical motion of
the cursor. The only purpose of this function is to produce
vscroll suitable for scrolling across large (relatively to the
window's height) images, where moving by display lines is not
appropriate.
* src/xdisp.c (Fdisplay__line_is_continued_p): New primitive.
* lisp/simple.el (line-move-partial): Call
'display--line-is-continued-p' to decide whether to leave it to
redisplay to scroll the window as appropriate. (Bug#48170)
* src/pgtkterm.c (pgtk_update_window_begin):
(pgtk_update_window_end):
(pgtk_redisplay_interface): Remove duplicate code and default to the
generic gui_update_window_begin/end implementations
Since tiling window managers may react allergically to resize
requests immediately following MapNotify events on X, make sure
that such requests are issued only when a new frame should not
become visible and a size has been explicitly requested for it.
* lisp/faces.el (x-create-frame-with-faces): Mark frame as
'was-invisible' if it should be initially invisible or iconified
and has its size specified explicitly.
* src/frame.c (make_frame): Initialize new frame's was_invisible
flag.
(Fframe__set_was_invisible): New internal function.
* src/frame.h (struct frame): Specify size of new_size_p slot.
New flag was_invisible.
* src/w32fns.c (Fx_create_frame)
* src/nsfns.m (Fx_create_frame)
* src/xfns.c (Fx_create_frame): Set new frame's was_invisible
flag.
* src/xterm.c (handle_one_xevent): Call xg_frame_set_char_size
after a PropertyNotify or MapNotify event only if F's
was_invisible flag was set.
In 66a36f1, I keeped passing text size by mistake.
* src/pgtkterm.c (pgtk_set_window_size): Don't use
FRAME_PIXEL_TO_TEXT_WIDTH/HEIGHT to pass pixel size.
* src/pgtkterm.c (pgtk_wait_for_map_event): New function to wait for
map events. The content is moved from pgtk_make_frame_visible.
(pgtk_make_frame_visible): Call pgtk_wait_for_map_event, instead of
process here directly.
(pgtk_make_frame_invisible): Call pgtk_wait_for_map_event to wait
for multiple map events.
* src/w32common.h: Rename OS_* to OS_SUBTYPE__*, as w32 API
headers started defining OS_NT, which breaks the use of the
enumeration. All users changed. (Bug#48303)
Previously, tooltips for menu items are below the menu.
To solve it, set tooltip text on menu items in advance, and entrust
gtk with handling tooltips.
* src/gtkutil.c (xg_show_tooltip): New function for pgtk to set
tooltip text on the widget.
(xg_hide_tooltip): New function for pgtk to clear tooltip text on the
widget.
(xg_create_frame_widgets): Don't set initial tooltips text, and don't
use qttip_cb.
(xg_create_frame_outer_widgets): Ditto.
(make_menu_item): Set tooltip text on the menu item.
* src/gtkutil.h: New declaration of xg_show_tooltip for pgtk.
* src/pgtkfns.c (Fx_show_tip): Calls xg_show_tooltip, and always ok.
This fixes bug #48249 and also a situation where, with recursive minibuffers
enabled and minibuffer-follows-selected-frame t, switching frames when a
minibuffer was open would leave the mini-window selected on the old frame.
* lisp/window.el (record-window-buffer): Add extra parameter DO-MINIBUF, and
amend the code such that minibuffers only get processed when that parameter is
non-nil.
* src/minibuf.c (zip_minibuffer_stacks, read_minibuf): Call
Qrecord_window_buffer with the new argument set to Qt.
(move_minibuffers_onto_frame): Set the selected window on the old frame when
this would otherwise remain the mini-window.
This should fix bug #47766.
* lisp/window.el (window-deletable-p): Add a quote where it was missing from
minibuffer-follows-selected-frame.
* src/frame.c (check_minibuf_window): Delete the function.
(delete_frame): In place of calling check_minibuf_window, call
move_minibuffers_onto_frame, possibly to move minibuffers onto the new current
frame.
(Fmake_frame_invisible, Ficonify_frame): Remove calls to check_minibuf_window.
* src/minibuf.c (Factive_minibuffer_window): Search the frames for the active
minibuffer rather than just assuming minibuf_window has been correctly
updated.
* src/xterm.c (handle_one_xevent): For GTK3 PropertyNotify and
MapNotify events explicitly request the stored frame sizes when
the frame changes from iconified to a non-hidden state
(Bug#24526). For Expose events do not change the frame's
visibility or iconified state. For FocusIn events on GTK3 do
not apply the fix for Bug#42655. The latter two changes are to
avoid that plain invisible frames get reported as iconified.
* src/indent.c (line_number_display_width): Make sure the selected
window's buffer is current before using display code on it:
redisplay assumes that the window's buffer is current at all
times. Reported by Evgeny Kurnevsky via lsp-mode's issue 1621,
https://github.com/emacs-lsp/lsp-mode/issues/1621.
Its only use was removed in the recent change of 2021-04-27 "Major
rewrite of adjust_frame_size", announced in the following thread:
https://lists.gnu.org/r/emacs-devel/2021-04/msg01162.html
* src/gtkutil.c [USE_GTK && !HAVE_GTK3] (gdk_window_get_geometry):
Remove unused macro to pacify -Wunused-macros build warning.
* src/xfns.c (x_set_tab_bar_lines): Call
x_change_tab_bar_height only if the number of tab bar lines
changed from or to zero (Bug#46827).
* src/xterm.c (x_make_frame_visible): Make frame_size_history
update less noisy by doing it only if the frame wasn't visible
before.
